Turn feedback into
reviewable evidence.
Import downloaded public-test reports, check their evidence and inspect differences by cohort. No report is automatically treated as validation.
Files stay in this browser tab. Nothing is uploaded or saved automatically. Export your reviewed collection before leaving; reopening the page starts empty. Review labels are your local assertions, not site certification. Do not import personal or vehicle-identifying information.
1. Import reports
Supports old and new public-test reports. Maximum 2 MB per file, 50 files per import and 2,000 records in this tab. Exact analytical duplicates are skipped; distinct reports are not verified unique vehicles. Models and snapshots must match this release.
Loading the frozen model…
2. Compare like with like
Positive differences mean the model is higher than the reported price. These are descriptive report differences, not independently validated errors. Date-mismatched prices mix model and market movements. Asking, sale and professional values are never pooled.
| Cohort | Reports | Median signed difference | Median absolute difference | Pass comparability checks |
|---|---|---|---|---|
| Import reports to begin. | ||||
Small cohorts remain descriptive. No minimum report count automatically establishes validity or permits calibration. There is no validated prediction interval, so interval coverage is not reported.
3. Record an evidence review
Comparability requires a supplied original price, same-vehicle evidence, ordinary condition, matching snapshot date, a source reference, declared non-use in development and a supported evidence review. Passing these checks still requires an independent sampling and validation design before any accuracy claim.